Abstract

Currently, the data processing approaches for 3D point cloud data are based on the topology of the data in reverse engineering. A novel data processing scheme for the 3D measurement results with scattered, unorganized and unordered 3D point data is presented. An improved algorithm, namely, unilateral left-side quartile method is put forward in this paper based on the frequency method and quartile in statistics. According to the presented algorithm, the point data are firstly portioned in space, based on frequency algorithm. Then the threshold of the noise data can be found from large number of data. K-nearest points can be found through the K-Dimensional binary search tree (K-D tree) established based on 3D point cloud, and several nearest border upon points (NBUP) are defined around each noise point. By means of the NBUP, the noise point data can be recovered. Finally, the effectiveness in reverse engineering of the proposed scheme is demonstrated via a testing result based on Handyscan 3D scanner.

Key words: Unilateral left-side quartile method, frequency method, unorganized point cloud, data preprocessing, K-D tree.